Summit offered research-based roadmap for law firms seeking to implement generative AI

I used to be honored to facilitate the “What’s Sizzling?” session for main giant legislation agency innovation professionals on the 2025 invitation-only Strategic Information & Innovation Authorized Leaders’ Summit in New York Metropolis on March 27, with Oz Benamram, the occasion’s co-founder and the previous chief information and innovation officer at Simpson Thacher & Bartlett.

In a mix of dialogue and exploration of survey analysis, the dialog served as a roadmap for legislation companies aiming to advance their practices into the long run.

The SKILLS survey interviewed roughly 100 professionals within the first quarter of 2025 relating to their use of varied synthetic intelligence functions. The analysis respondents primarily got here from giant legislation companies, though companies of all sizes that emphasize innovation have been inspired to take part. The overwhelming majority (70) have been primarily based in the US, with 9 from the UK, eight from Canada and 13 from worldwide legislation companies, primarily cut up between New York and London headquarters and varied different nations. Finishing the survey was a requirement to attend SKILLS.

Massive legislation companies are deploying generative AI

The report recognized 22 use instances and 180 options. On common, the legislation agency leaders that responded have 18 lively generative AI options, six in a pilot part and 22 into account. The commonest use instances included contract drafting, time entry, due diligence and authorized analysis.

One foundational precept resonating with the group is the management important for driving change, and contributors provided a number of suggestions to realize help. First, efficient change administration calls for constant communication from the highest down all through the group. Second, innovation leaders who report on to an government, together with the managing associate, might wield extra affect in implementing change. Third, cross-functional management can improve a group’s worth for particular tasks.

Almost all SKILLS legislation companies have AI use insurance policies

Almost all legislation companies that responded (99%) have an AI use coverage. Nevertheless, fewer companies have developed an AI technique (92%) or created an AI process power (87%). These outcomes underscore the numerous significance that legislation agency leaders place on tactical AI steerage, efficient deployment protocols and collaborative expertise.

Additionally they function very important benchmarks for innovation groups devoted to laying the groundwork for brand spanking new initiatives. With out insurance policies, methods and expert groups, firms might wrestle to realize the important belief of their inside stakeholders and shoppers wanted to succeed with generative AI. That is significantly necessary as many authorized professionals nonetheless concern and stay unfamiliar with a number of AI functions utilized by the respondents.

Growing an AI roadmap requires management and a framework

In the course of the dialogue, attendees really useful that insurance policies governing the usage of synthetic intelligence embrace provisions for ethics, restrictions on firm-approved merchandise, the importance of human involvement and a framework for threat evaluation. Firms must also improve their efforts by collaborating with exterior suppliers to complement restricted in-house assets.

Whereas the adoption of generative AI at some companies is robust and sustained, the viewers agreed that no clear demographic patterns exist. One attendee famous that the extra senior an expert is, the higher worth they will derive from utilizing an AI platform by asking “smarter questions,” given their depth of data and skill so as to add substance to a immediate.

Innovation groups are adapting to the shift from instruments to transformation

Unsurprisingly, the innovation groups on the responding legislation companies maintain major accountability for implementing the companies’ AI methods (59%), intently adopted by information administration (44%) and IT (43%), each of which share an analogous degree of oversight.

Whereas the innovation group’s involvement has remained secure over the previous 12 months (59% in 2025 in comparison with 62% in 2024), information administration (44% in 2025 in comparison with 63% in 2024) and IT (43% in 2025 in comparison with 74% in 2024) have seen modifications. Moreover, 1 / 4 of respondents in 2024 indicated that different departments supplied enter, and 15% relied on exterior help for his or her AI initiatives.

Particular legislation agency developments have influenced which groups are liable for sure actions. The dialogue amongst legislation agency leaders relating to the deployment of synthetic intelligence has rapidly transitioned from expertise to transformation. Legislation companies are more and more reexamining their complete suite of functions and evaluating the standard and integrity of their knowledge to make sure that any giant language fashions they use can ship one of the best outcomes.

In 2025, this can focus much less on expertise and extra on redefining the legislation agency’s technique for maximizing the worth of its info. Consequently, the innovation, IT and information administration groups will work collectively to navigate AI’s transformative influence on the agency.

Many organizations lacked readability a 12 months in the past as legislation companies have been nonetheless assessing potential functions and use instances. The fast tempo of AI growth and the insatiable demand for its availability necessitates a cross-functional and interdisciplinary group to drive the agency ahead, with a delegated ambassador coordinating its proliferation. That diplomat is the pinnacle of innovation at many companies, whose obligations are broader and deeper than in 2024, but the diplomatic corps stays multifaceted.

Particularly, the function of IT groups in legislation companies has grow to be considerably extra distinguished because the pandemic. As budgets have elevated, cybersecurity has grow to be important, and progress is intently aligned with expertise. This place emphasizes holistic operational excellence slightly than merely digital transformation.

One other necessary development to notice is the growth of data administration to incorporate innovation, authorized operations and synthetic intelligence. Right this moment, legislation companies steadily recruit expertise from information administration and both assign obligations for innovation or promote a KM chief to a broader innovation-focused function. Whereas solely 35% of respondents work at companies which have appointed a “Chief Synthetic Intelligence Officer” or created an analogous management place, this can probably change and affect who addresses or navigates these challenges.

Whereas the information signifies that innovation groups have gotten the first AI stakeholders, the possession construction stays extra nuanced because of overlapping expertise and expertise. Moreover, extra professionals use the time period “innovation” of their titles.

The vast majority of legislation companies have deployed generative AI

Greater than two-thirds of contributors (73%) have applied an inside generative AI answer—both developed an AI chatbot internally or deployed one by means of a safe API. Sixteen % reported that they constructed or commissioned a customized inside language mannequin. Though unusual, virtually 1 / 4 (24%) have created client-facing or revenue-generating merchandise utilizing generative AI. This development seems to reflect the actions of legislation companies a number of years in the past, which initially developed cellular apps for inside use and later started providing extra complete variations to shoppers.

One participant prompt creating a completely accessible library for generative AI prompts and project-specific suggestions, akin to Netflix for queries.

OpenAI’s GPT is the main LLM in legislation companies, however the generative AI help market is fragmented

Among the many 73% of respondents who’ve applied a generative AI answer, about three-quarters (76%) use OpenAI’s GPT as their foundational LLM mannequin. Anthropic’s Claude ranks second with 19%, whereas Google‘s Gemini ranks third with 15%. Apparently, 3% reported growing with China’s DeepSeek.

A 3rd (33%) reported utilizing an exterior supplier to assist them construct their language mannequin; nonetheless, the market stays fragmented. Respondents talked about about 15 utilized AI companions. Though consolidation is predicted, the present demand will probably drive short-term progress slightly than contraction within the supplier market.

Encouraging the adoption of generative AI stays difficult

Whereas 73% of respondents indicated that their companies make the most of a authorized AI assistant or chatbot, 62% have adopted the applying throughout your entire agency or a considerable subset for testing functions.

Adoption charges have remained low, with solely 22% reporting a 50% to 100% adoption charge. This low adoption degree is one cause sure groups stopped utilizing particular instruments. Different elements embrace cost-effectiveness and the provision of higher alternate options. Because the market grows and the boundaries to entry diminish because of developments in strong AI fashions, competitors will proceed to accentuate.

To make sure success, the attendees agreed that legislation companies ought to start their initiatives with people who’re keen and passionate, particularly since adoption is rising inside the present consumer base slightly than from a surge in new contributors in any given challenge.

Moreover, when evaluating expertise and processes, exploring markets past the US, particularly in Europe, could be advantageous for producing recent concepts and buying an intensive understanding of the evolving authorized trade.

In some companies, discussions about generative AI use instances are reigniting curiosity in doc automation. It’s experiencing a renaissance, as it might present a simpler and cost-efficient answer to the challenges that some professionals count on generative AI to handle.

AI authorized assistants and chatbots are main use instances

The analysis indicated that an AI authorized assistant or chatbot was the preferred use case. Different vital functions included knowledge extraction, summarization, authorized drafting, contract evaluation and evaluation, proofreading, analysis and looking out. In distinction, the least cited functions have been patent and IP drafting, litigation administration, automation of discovery responses, pitches and proposals, and compliance.

When contemplating a brand new utility, “attempt to keep away from signing a long-term settlement,” suggested an viewers member, who added, “Value renewals are additionally absurd, and an analogous software is usually out there at half the fee with higher performance.” A peer provided, “Functions and not using a distinctive worth proposition, i.e., a moat defending their aggressive benefit, might not be value a long-term funding.” Lastly, one participant cautioned, “Don’t attempt to study all the pieces as a result of the expertise will proceed to evolve and solely get quicker.”

Ari Kaplan is a lawyer and authorized trade analyst who publishes benchmarking studies on authorized expertise developments and is an skilled webinar and convention facilitator.

This column displays the opinions of the creator and never essentially the views of the ABA Journal—or the American Bar Affiliation.